Research map / Label-efficient and robust learning / Semi-supervised learning
When Active Learning Meets Implicit Semantic Data Augmentation
Zhuangzhuang Chen, Jin Zhang, Pan Wang, Jie Chen, Jianqiang Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add EPMCM selection plus ISDA (DAST-AL) versus random with ISDA (ablation) | Ran+ · CIFAR-10 (ResNet-18), 10% labeled | accuracy: 80.76 -> 82.98 | Improved |
| add EPMCM selection plus ISDA versus random with ISDA (ablation) | Ran+ · CIFAR-10 (ResNet-18), 35% labeled | accuracy: 91.15 -> 93.32 | Improved |
| add EPMCM selection plus ISDA versus max-probability with ISDA (ablation) | Maxp+ · CIFAR-10 (ResNet-18), 20% labeled | accuracy: 81.72 -> 90.42 | Improved |
| add ISDA to random selection (ablation) | Ran · CIFAR-10 (ResNet-18), 10% labeled | accuracy: 80.06 -> 80.76 | Improved |
| add ISDA to random selection (ablation) | Ran · CIFAR-10 (ResNet-18), 40% labeled | accuracy: 91.21 -> 91.15 | Hurt |
| add ISDA to random selection at 5% labeled (ablation) | Ran · CIFAR-10 (ResNet-18), 5% labeled | accuracy: 67.13 -> 69.27 | Improved |
| add ISDA to max-probability selection (ablation) | Maxp · CIFAR-10 (ResNet-18), 10% labeled | accuracy: 76.37 -> 76.35 | No clear effect |
| use max-probability selection instead of random (ablation) | Ran · CIFAR-10 (ResNet-18), 35% labeled | accuracy: 91.21 -> 84.60 | Hurt |
| increase number of explicit augmented samples from 100 to 100000 in the upper-bound gap (ablation) | EPMCM upper bound · CIFAR-10 ResNet18 gap | gap: 0.85 ± 0.49 -> 0.11 ± 0.07 | Improved |
| use DAST-AL instead of ADS for first-iteration time | ADS · CIFAR-10 timing | total time (min): 44.91 -> 2.26 | Improved |
| use DAST-AL instead of SRAAL for AL iteration time | SRAAL · CIFAR-10 timing | total time (min): 298.9 -> 2.26 | Improved |
| use DAST-AL instead of ADS at 2% labeled data | ADS · CIFAR-10 VGG-16 (2% labeled) | accuracy | Hurt |
Closest papers by meaning
- Consistency-based Semi-supervised Active Learning: Towards Minimizing Labeling Cost (ECCV 2020 · similarity 0.88)
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 2024 · similarity 0.87)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
- Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning (CVPR 2025 · similarity 0.86)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.85)
- Omni-Query Active Learning for Source-Free Domain Adaptive Cross-Modality 3D Semantic Segmentation (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
